The performance and flow characteristics of a gas turbine combustor dump diffuser

Transcription

1 Loughborough University Institutional Repository The performance and flow characteristics of a gas turbine combustor dump diffuser This item was submitted to Loughborough University's Institutional Repository by the/an author. Additional Information: A Master's Dissertation, submitted in partial fulfilment of the requirements of the award of the Master Of Science degree of Loughborough University. Metadata Record: Publisher: c Kevin Andrew Goom Please cite the published version.

2 This item was submitted to Loughborough University as a Masters thesis by the author and is made available in the Institutional Repository ( under the following Creative Commons Licence conditions. For the full text of this licence, please go to:

3 LOUGHBOROUGH UNIVERSITY OF TE<;:HNOLOGY LIBRARY AUTHOR..._...._100 ~ -r-.. Ib c:?.~.!... ~:.-... Q~1 ~~r;"./9_i._. VOL NO. CLASS MARK \ 8 MAl 1~98 23 DCT 200.'

4

5 'e. fk/~ 3b(

6 THE PERFORMANCE AND FLOI'/ CHARACTERISTICS OF A GAS TURBINE COf1BUSTOR DmlP DIFFUSER by KEVIN ANDREW GOOM A MASTER'S THESIS Submitted for the award of Master of Science of the Loughbo::-ough University of Technology December 1974 Supervisor: Dr. S.J. Stevens, Department of Transport Technology. C by Kevin Andrew Goom, 1974.

7 Lou9hto,cu;~ llniverlity' of T::rhr, "':::oy L!;"~ry I ~ --~ ~:':.'-]~J-L. l C:i~S

8 ( i) SUMMARY Low speed tests have been carried out on a branched diffuser, geometrically similar to that employed in some gas turbine combustion systems. The fully annular test rig consisted of a straight walled, axisymmetric pre-diffuser, exhausting into a sudden expansion, the flow then being divided into two separate streams by a bluff body simulating the combustion chamber situated on the same centre-line as the pre-diffuser. The overall area ratio was maintained at 2.0, and tests vjere conducted with fully developed entry flo\~ for a range of pre-diffusers of 12 0 included angle, the design value of the ratio of mass flows in the inner and outer annuli surrounding the flame tube being fixed at 1.2. Further tests were conducted using a single pre-diffuser, a distorted entry velocity profile, and a design flow division around the flame tube of The influence of variation of.the division of flow around the bluff body, and the axial location of the bluff body, vjere investigated at each of the conditions. cited above. The performance in terms of total pressure loss and static pressure recovery was evaluated for the system as a whole, and for the regions between measurement stations. It was found that optimum performance both overall, and for the pre-diffuser alone, occurred at a flow division close to the design value of 1.20, and corresponded to a symmetric pre-diffuser outlet velocity and static pressure distribution. By bringing the bluff body closer to the pre-diffuser, at a flo\~ division close to the design value, it was found that considerable improvement of pre-diffuser pressure

9 (H) recovery and outlet flow uniformity was achieved. However, when the distance between bluff body and pre-diffuser was small, considerable loss and pre-diffuser outlet flow distortion resulted at off design conditions. Increase of pre-diffuser area ratio had the effect of increasing loss within the pre-diffuser, but reducing loss downstream. For each flow split and bluff body location, an optimum pre-diffuser area ratio existed. In view of the increase of turbulence associated with distortion of the entry profile, it was not possible to isolate the influence of velocity profile distortion alone. However, the overall effect was to increase loss at all operating conditions. The results of this investigation, and earlier work, emphasize the need to match the system geometry and design flow division.

10 (iii) ACKNOVILEDGEMENTS The work described in this thesis was carried out in I the Department of Transport Technology at Loughborough University of Technology under the financial support of the Ministry of Defence. The author gratefully acknowledges both of these bodies for providing the opportunity to conduct this research. Thanks are extended to Dr. S. J. Stevens who supervised this work, and devoted considerable time and effort providing invaluable assistance. The author also wishes to thank Mr. W. H. Brooks for his excellent work in connection with the construction of the test rig and instrumentation. Miss V. M. Johnson is also acknowledged for her excellent typing of this thesis.

11 (iv) LIST OF CONTENTS Section Title Summary Acknowledgements List of Contents List of Figures and Tables Nomenclature (i) (ii) (iv) (vii) (xii) 1. INTRODUCTION The Characteristics of a Diffuser The Classification of Diffusers The Use of Diffusers in Gas Turbine Combustion Systems Performance Evaluation.The Operation ofa Combustion Chamber Dump Diffusing System Review of Previous Work The Choice of System to be Investigated The Scope and Aims of the Investigation THE EXPERIMENTAL FACILITY The Construction of the Test Rig The Test Rig Variables Instr)lmentation EXPERIMENTATION 3.1 Scope of Tests 34

12 (v) 3.2 Entry Conditions Experimental Technique Accuracy PRESENTATION AND DISCUSSION OF RESULTS 4.1 Overall Performance ( 1-4) Pre-Diffuser Performance (1-2) Dump and Head Region (2-3) Settling Length (3-4) Division of Loss Within the System CONCLUSIONS AND RECOMMENDATIONS 5.1 Conclusions Recommendations for Future Work 67 REFERENCES 69

13 (vi) APPENDICES Appendix Title 1. DATE ANALYSIS A.l.l A.l.2 Data Preparation Computer Program PRE-DIFFUSER OUTLET PROFILES HEAD STATION (3) PROFILES SETTLING LENGTH PROFILES PERFORMANCE PARAMETERS BOUNDARY LAYER PARAMETERS 227..

14 (vii) LIST OF FIGURES AND TABLES Figure/Table No. Title. Page Two Types of Combustion Chamber Diffuser Systems in Current Use Variation of Overall Ideal Pressure Recoveries and Effective Area Ratios With Flow Split Relationship Between Kinetic Energy Flux Coefficient and Blockage Fraction Comparison of Pre-Diffuser and Isolated Diffuser Pressure Recoveries Influence of Inlet Blockage on Diffuser Performance after Sovran and Klomp(2) Variation of Diffuser Effectiveness wi th Inlet Blockage Fraction Influence of Entry Swirl on Loss Coefficient for Constant Inner Core Annular Diffusers after Gurevich(5) Total Pressure Contours in Annular Diffusers Layout of Test Facility Experimental Facility 78 "

15 (viii) Test Rig and Pre-Diffuser Geometries Settling Length Throttle and Traverse Mechanism Pre-Diffuser Geometries in Relation to Performance Chart of Sovran and Klomp(2) Location of Instrumentation Details of Head Rakes Location of Static Pressure Tappings Settling Length Traverse Probe Traverse Mechanism and Pitot Probe Wedge Static Probe Summary of Tests Conducted Entry Velocity Profiles Details of the System Used to Generate a Distorted Entry Velocity Profile Shear Stress and Velocity Distributions at Inlet for Various Ring Positions Calibration of Pressure Probes at Incidence 89

16 (ix) 4.1.1/3 Overall Performance - 90 Pre-Diffuser Nos. 1/ Variation of Maximum Pressure 93 Recovery and Minimum Loss with Overall Length Example of the Variation of 94 Inner and Outer Overall Pressure Recoveries with Flow Split Variation of Inner and OUter 94 P):'essure Recoveries with Overall Length at Optimum Flow Splits Design Chart for Maximum Pressure 95 Recovery (S = ) Design Chart for Maximum Pressure Recovery (S = ) Design Chart for Minimum Loss Examples of Pre-Diffuser and Overall 98 Performance Variation with S, Land D General Flow Stability Overall Performance - Pre-Diffuser 100 No Example of the Variation of Pre-Diffuser 101 Outlet Profiles with Dump Gap

17 (xl Pre-Diffuser Outlet Profiles for 102 Test Series /5 Pre-Diffuser Performance 103 Pre-Diffuser Nos. 1/ /8 Pre-Diffuser Outlet Flow Distortion 106 Parameters - Pre-Diffuser Nos. 1/ Typical Variation of Pre-Diffuser 109 Pressure Recovery with Dump Gap The Reduction of Pre-Diffuser 110 Pressure Recovery Attributable to Loss Pre-Diffuser Separation Limits Pre-Diffuser Performance 112 Pre-Diffuser No Pre-Diffuser Outlet Kinetic Energy 113 Flux Coefficient Variation Pre-Diffuser Outlet Profiles for 114 Test Series Performance Parameters for Region Pre-Diffuser No. 2, S = Head Inner Profiles for Test 116 Series Head Outer Profiles for Test 117 Series 2-05.'

18 (xi) Some Factors Influencing Performance Between Stations 2 and Key to Combustion Chamber Static Pressure Distributions /8 Combustion Chamber Static Pressure Distribution for Diffusers 1/ Performance Characteristics for Region Head Station (3i) Velocity Profile Characteristics - Pre-Diffuser No '4.4.3 Settling Length Velocity Profiles for Tests Series /3 Division of Loss Throughout the System - Pre-Diffuser Nos. 1/3 126 A.l.l Flow Diagram for Performance Analysis Program 131 A.l.2 List of Principal Symbols in Analysis Program 132 A.l.3 Listing of Performance Analysis Program 134

19 (xii) NOMENCLATURE Cross-sectional area Blocked area Area ratio ARe B Cp Cp' Cp Cp D E H h Lov m p p q R Re Effective area ratio of a branched system Blocked area fraction Static pressure recovery Ideal static pressure recovery Locus of maximum pressure recovery for a given diffuser length Locus of maximum pressure recovery for a given diffuser area ratio Dump gap Effective area fraction Boundary layer shape parameter Annulus height 'Overall' system length (L + D) Length of a simple diffuser Mass flow Total pressure Static pressure Dynamic head Radius Reynold number Radius of combustion chamber head Mean annulus radius Velocity profile radial distortion factor S Flow split ratio Design flow split

20 (xiii) u u, u v y, Maximum velocity Local velocity Fluctuating component of velocity in 'x' direction (axial) Fluctuating component of velocity in 'y' direction (radial) Distance from wall Kinetic energy flux coefficient Momentum flux coefficient ~ e /. v t f o Stability parameter Boundary layer displacement thickness Diffuser effectiveness Inclination of diffuser to axial direction Boundary layer momentum thickness Total pressure loss coefficient Kinematic Viscosity Derisity Diffuser wall angle Superscripts Area weighted mean.v' Mass weighted mean Subscripts H i min max Referred to combustion chamber head Inner wall or annulus Minimum value Maximum value

21 (xiv) m o w II Value at point of maximum velocity Outer wall or annulus Value at wall Diffuser or diffuser system inlet station Diffuser or pre-diffuser outlet station Combustion chamber head station Settling length station Quantity based on a two-dimensional definition

22 - 1 - SECTION 1 INTRODUCTION 1.1 THE CHARACTERISTICS OF A DIFFUSER A diffuser is basically a duct, the cross-sectional area of which increases in the direction of flow. In the absence of transfer of fluid across the duct walls, the mean velocity of the flow must decrease, this generally being accompanied by a rise of static pressure. Hence, a diffuser may be classed as a device which converts kinetic into potential energy. In the adverse pressure grudient through a diffuser, boundary layers will tend to grow. In some cases, this can lead to separation, which is normally undesirable for the following reasons, although devices such as vortex generators or boundary layer bleed may be used to minimise this effect. (i) The increased total pressure loss associated with separation. (ii) The point at which separation occurs is often unpredictable, and furthermore unstable. The resultant velocity and pressure fluctuations can have serious adverse effects on adjacent components. (iii) The static pressure rise is reduced, since the main flow does not occupy the whole of the duct, so decreusing the effective area ratio of the diffuser. Even in the absence of separation, the boundary layer growth is normally undesirable, since it produces a velocity profile,which has an increase of axial non-uniformity, and

23 - 2 - since this has a greater kinetic energy than that of a uniform profile, the effect is to reduce the amount of diffusion possible. 1.2 THE CLASSIFICATION OF DIFFUSERS In order to fully describe a diffusing system, it is necessary to define: (i) The basic type (two-dimensional, conical, annular) (H) (iii) The wall'shape (straight or contoured) The dimensions Within the scope of this work only straight walled annular diffusers are considered. The following parameters 1 have been found useful in classi:fying this type: " ~-.j a ~=-==--..::-~----l:e------b RI RIo Rli (U The ratio of inlet annulus height to mean radius

24 - 3 - (ii) The ratio of inlet annulus height to mean length L/hl (Hi) The inclination, ~, of the entry axis (a-b) to the diffuser axis (a-a) (iv) The diffuser included angle 20 Hence four geometric variables are necessary to fully define a straight walled annular diffuser. It is often useful to know both the rate and amount of diffusion being attempted. The former can be expressed as the geometric area ratio of the system AR = x sin E.] reducing to AR = ~+2 tan ~ for non-inclined diffusers ( E =0) The overall rate of diffusion is generally expressed in terms of the amount of diffusion being attempted per unit length, which can be written non-dimensionally as: h = 2 tan 0 + (_1) sine. tan 0 sin Rl or for non-inclined systems, = 2 tan THE USE OF DIFFUSERS IN GAS TURBINE COMBUSTION SYSTEMS.' Typically, air from a gas turbine compressor emerges at a Mach Number of about 0.3. Since kerosene fuel has a low flame speed, a certain proportion of the flow must be

25 - 4 - diffused before stable combustion can be attempted. Two typical system used to achieve this are shown in Figure Typically, having passed through the pre-diffuser, 18% of the air enters the primary zone prior to which, suitable amounts of mixing and swirl are introduced, which sets up the desired turbulent, low velocity flow conditions necessary to promote stable combustion. A further 10% enters through the flame tube walls to complete the combustion process in the secondary zone, the remaining air being used for diluting the combustion products, and cooling the chamber walls. However, the current trend, with the requirement for low pollution engines, is to increase the percentage of flow entering directly into the primary zone, in an attempt to ensure complete combustion. This has only been made possible by the introduction of more sophisticated wall cooling techniques, \1hich ensure that the flame tube temperature is kept within permitted limits, despite the reduction of air available for this purpose. The combustion and cooling processes must be effective over the whole operating range of the engine, or phenomena such as 'hot spots', and incomplete combustion may occur, giving rise to inefficiency, pollution, a poor turbine entry temperature distribution, and even mechanical damage. Hence the system must operate in a predictable, stable and uniform manner at all engine running conditions. One of the major sources of flow non-uniformity is separation, since it will introduce instability, and circumferential flow non-uniformity, the latter in view of the uncertainty of the location of separation inception. l'his

26 - 5 - is a particular problem with the faired system, since boundary layers are encouraged to grow on both combustion chamber and splitter walls due to the adverse pressure gradient cre- I ated by the diffusion process. Furthermore, the pressure gradient developed as a result of turning the flow induces the boundary layer on one wall to grow even more rapidly. Separation can normally be avoided by making a system of sufficient length, but this is often undesirable, particu. larly in aircraft applications, in view of the need to keep engine weight, and length to a minimum. The dump system is an attempt to create a short, stable system by fixing the separation at an abrupt expansion, and replacing the splitter plates with a blunt nosed head. Since the flow is now split by this blunt body, this system is less sensitive to a change of flovj division around the head than the faired system, the splitter plates of which must operate at incidence when the division of flow is other than the design value. In modern, high by-pass ratio gas turbines, the airflow through the gas generator is often quite low, and the size of the combustion system also correspondingly low. Small dimensional inaccuracies arising during manufacture, or distortion in operation can therefore have a significant effect on the flow within the combustion system. This is a particular problem with the faired system in which three small area ducts exist downstream of the pre-diffuser, this being replaced by a far greater area in the sudden expansion of the dump system. Further features relating to the operation of a dump

27 - 6 - diffuser are discussed in Section PERFORMANCE EVALUATION In any diffuser work, it is necessary to have a set of parameters which may be evaluated in order to indicate how well a system is performing. The most important are: (i) The amount of diffusion which has been obtained (i.e. the static pressure rise) (ii) The total pressure loss incurred (iii) The degree of flow non-uniformity, both axially and radially, and possibly in addition, circumferentially. Hence any system of performance presentation must include parameters which give a useful measure of these quantities, presented in non-dimensional form AVERAGING METHODS In general, the properties of a fluid at any given position in a duct will be distributed non-uniformly. Since it is usually necessary to calculate overall changes between various duct positions, it is necessary to introduce as averaging technique which converts the flow into a one-dimensional equivalent. follows: Several such methods can be used, and are as (i) Area weighting (ii) Mass derivation (iii) Mass weighting (iv) Momentum mixed weighting )

28 - 7 - Ideally, the one-dimensional equivalent is required to have identical properties to the flow that it represents i.e. the same mass, momentum and energy fluxes. Since a diffuser is primarily a device for converting energy from one form to another, the energy criterion must be satisfied as a prime objective, hence ruling out (i) and (ii) above. Mass weighting has been chosen for the purposes of this work since it is in more common general usage, and is consistent with the averaging method suggested by Livesey(14) as being the correct one to give meaningful equivalents of non-uniform parameters. In all following derivations, incompressibility has been assumed, justified in view of the fact that the Mach number never exceeds about 0.1. The mass weighted mean of a parameter X is defined as:... JXmdm X = vlithin this system, it is convenient to revert to an area weighted mean in the single case of velocity. In this case, the area weighted mean of velocity, u, is defined as: u = J -- u da A A Total mass flow, m, is given by: m = A f f uda = f /UdA substituting from yields: 1.4.3

29 - 8 - noting that dm = f uda, the definition of mass weighted mean can be re-written as: X '" = A J X~dA ua MASS WEIGHTED PARAMETERS At this stage, it is convenient to define a parameter which gives a measure of the extra energy associated with a distorted velocity profile, compared with that of a uniform profile of the same mass flow. The kinetic energy flux coefficient is useful in this respect, and is defined as: 0( = (kinetic energy of flow) (kinetic Energy of flow with the same mass flux, but uniform velocity profile) 0( = -2 2 m u = J A momentum coefficient may be defined in a similar way: DYANMIC HEAD The mass weighted mean of dynamic head, q, is given by: '" q = = substituting from '. AJ 3 1. f' -=.u----::d::...:.:.a 2 U A 1.4.7

30 - 9 - '" q = I 1 "2 f 0<. A U 3 A U = 1f -2 0(- u 2 Hence, mass weighted mean dynamic head may be found directly by integration (1.4.7), or indirectly from the kinetic energy flux coefficient, and an area weighted mean velocity (1.4.8) PERFORMANCE PARAMETERS SIMPLE DIFFUSERS The basic definition of loss and pressure recovery coefficients are as follows: = \1\ C p1-2 = Writing an energy balance between stations 1 and 2: if no fluid is transfered across the duct walls, then or- '" P1 + q1 = '" '" '" '" P2 + q2 (p 1 - P2) dividing by ql and re-grouping: '" '" '" '" '" P 1 P 2 q2 _ P2 P1 = 1 - ( ) '" '" '" ql q1 ql

31 substituting from 1.4.8, 9, 10: for continuity also A2 Al = AR l _ 2 ~1-2 '" 1-0(2 "<1 1 2 (AR) - '" c pl-2 Hence loss coefficient may be found by direct integration of total pre~sures (1.4.9), or indirectly from The latter has been found more convenient for the purposes of this investigation. The ideal pressure recovery may be found from Equation by assuming no loss ( A _ = 0), and a uniform velocity l 2 profile at station 2 ( 0<2 = 1.0) v-' c pl-2 = (AR ) 0( BRANCHED DIFFUSERS ~ INNER FLOW FIELD ~ OUTER FLOW FIELD /1/) o::j W -l/ z:j ZZ - z / / <! DIVIDING STREAMLINE

32 Two definitions are available to express individual flow fluid performance. They differ only in the reference dynamic head used to non-dimensionalise the equation, namely I mean entry or individual flow field entry dynamic head, the latter hereafter being called 'split' entry conditions. r(ef erencing to mean entry conditions: '" '" PH - P4i V" ( A 1 _ 4 \ = V' q1 '" '" P4i - P1'.~ (C'" p1-4\ = '" q Referencing to split entry conditions: '" V' V' P 1i - P4i ( f.. 1 _ 4 \ = V' qu VI '" "" P4i - PH (C 1 4)' = p - ~ '" qu Writing the energy equation for the inner portion of the divided flow: V" '" "'"' \of' m 1i (P1i + q1i) = m 1i (P4i + q4i) + m 1i P 1i - m 4i P4i dividing by putting m 4i = m 1i, and re-grouping = 1 '" '" P 4i PH 'El 1i u 4 2 (-~) Uu substituting from , 17 0(4i ~i '" (C 1 4)' p - ~ 14.18

33 (coefficients referred to split entry mass weighted mean dynamic head) Similarly, using the definitions of , 15, it can be shown that: u 1, 2 (-~) u 1 u 4, 2 (-,2:.) (C'" 1 4)' p - ~ (coefficients based on mean entry mass weighted mean dynamic head) It is of interest to note, that by writing the energy equation for both parts of the divided flow: we obtain:- and substituting for the definition given in , 17 (i.e. mean inlet reference), Hence when using definitions based on mean entry conditions, the overall loss coefficient, and similarly pressure recovery, can be found by the sum of the mass weighted means of the individual flow field components. For calculation purposes it may be convenient to use the flo\~ split ratio, S = m40/m Ai, when determining overall performance parameters. By substitution of into , and noting that (AR l _ 4 )i = A4i/A11 it can be shown that:

34 - '13-1 (_1_)3-0<1 1+$ v- C pl Pressure recovery can also be expressed in this form as: The ideal pressure recovery follows from by '" putting ~1-4 = 0, and ~i = ~o = 1 = 1 1_ (-1.-) 3 ( 0<.1 1+$ An effective area ratio may now be introduced, defined as the area ratio of a simple diffuser in which the amount of diffusion being attempted is the same as that of the branched system. (i.e. the ideal pressure recovery is identical) From and it follows that: =.( )-2 3( )-2 AR l _ 4 i + 5 AR l _ The effective area ratio equals the geometric area ratio at only one value of flow split. This is when the mean velocities in the two divided streams are equal (u 4i = U 4o ),

35 and corresponds to a point at which the ideal pressure recovery is,a maximum (see Fig ). This value of flow split is termed the design value, s', and is given by: s = LOCAL PERFORMANCE PARAMETERS It is often desirable to divide a flow into sections, and view the performance of each part individually. This section concerns parameters which may be defined to satisfy this requirement. Firstly, we may use the overall inlet conditions as a reference, yielding for example: This system enables overall performance parameters to be obtained as the sum of individual components, e.g.: V" V' '" = C _ + C _ + C _ pl 2 p2 3 p3 4 However, it may be preferable to use local entry conditions as a non-dimensionalising factor, e.g Additionally, local performance parameters may be defined for individual (i.e. inner or outer) flow fields. In this case, a mean or split entry dynamic head may be used in the definition e.g.

36 - '1.5 - '" '" V' P3i - P2i (C p2-3)i = '" q1 using overall entry conditions as a '" '" reference '" P3i - P2' ~ or (C 2 3)' = p - ~ '" q1i '" '" '" P3i P2i (C 2 3). = p - ~ "'" q2 using local entry conditions as a '" '" reference '" P3i - P2i. (C 23)' = P - ~ '" q2i BOUNDARY LAYER PARAMETERS In the present work, boundary layer parameters have been found useful as a means of quantitatively describing the development of velocity profiles throughout the system. Generally recognised definitions have been used: non-dimensional displacement thickness, R -R, wo w~ = RRmJ (1 - ~) ~w dr w ~R -R,) I \ wo Wl non-dimensional momentum thickness, e R -R, wo w~ = :1 IcR -R,) I \ wo w~ Shape factor, H = where R, w~ radius of the wall to \ojhich the boundary

37 layer relates = R = wo (R -R,) = wo w~ radius to which the boundary layer under consideration extends radius to which the annulus extends is the annulus height, and is used to obtain non-dimensional parameters U = maximum velocity i.e. at Rm Parameters evaluated in the region of the combustion chambers head (station 3), require further clarification. PLANE OF HEAD MEASUREMENT "" I : R,,, /\ VORTEX LIMIT ",~I_- I The annulus height h3 is used to non-dimensionalise, but integration is proceeded only up to the start of the vortex region.

38 VELOCITY PROFILE PARAMETERS In diffusers, it is often useful to be able to quantify the change of shape of velocity profiles as the flow progresses through the duct. Two parameters are in general use which give a measure of the axial non-uniformity. Firstly, the kinetic energy flux coefficient, as defined in Equation may be used, since the kinetic energy of a flow rises with increasing axial non-uniformity. Secondly, a blocked area concept may be used, which indicates the surfeit of area required to transmi~ a certain mass of fluid over that required to transmit the same mass -of fluid, in a flow of uniform velocity, equal to the maximum velocity of the non-uniform flow. The blocked area, AB' is given by: = A f (1 - ~) da - = A(l - ~) The blocked area fraction, B, and effective area, E, are then given by: B = (1 - ~) = These parameters are simplex to calculate than ol, since detailed knowledge of the velocity profile is not required, only mass flow rate, duct area, and the maximum velocity. However, for the purposes of this investigation, it has been found more convenient to use 0( as a measure of axial distortion. The relationship between 0(2 and B 2, taken from pre-diffuser outlet test data, is given in

39 Figure In order to provide an assessment of radial asymmetry, a parameter is required which is independent of axial nonuniformity. The position of maximum velocity could be used, but this point is often ill defined, particularly in fairly uniform profiles. The boundary layer parameter, di., is far less sensitive to accurate determination of the maximum velocity point, in view of the zero near the point. term which tends to A parameter, hereafter termed the radial distortion factor, R D, can then be defined as: J~ - J O ~ 0 = cf~ + J~ Where i and 0 refer to the two regions of the profile, divided about the point OL maximum velocity. For a symmetric velocity profile, RD = 0, but becomes negative for a profile distorted towards the inner duct wall, and positive when distorted towards the outer wall THE OPERATION OF A COMBUSTION CHAMBER DUMP DIFFUSING SYSTEM PRE-DIFFUSER DUM P REGION HEAD REG ION SETTLING LENGH '-.'

40 The overall area ratio of a dump diffuser is of prime importance, since it governs the amount of diffusion being attempted. It is largely fixed by the Mach number at compressor exit, and the necessity to obtain a sufficiently high static pressure in the settling length annuli to ensure adequate penetration of the flow into the flame tube walls. Considerable significance is also attached to the pre-diffuser area ratio, since this, in part, dictates. the amount of diffusion which must subsequently occur further downstream. The purpose of pre-diffusion is to attempt to minimise loss by reducing diffusion in the potentially high loss region of the turning flow around the combustion chamber head. Consideration must however be given to the length of the pre-diffuser, and the non-uniformity of flow produced by it, with particular regard to separation. The combustion chamber head, hereafter referred to simply as the head, can however improve the flow conditions at exit from the pre-diffuser. A region of high static pressure, centred around the stagnation point, must exist on the head, creating, particularly at small dump gaps, a non-uniform static pressure distribution across the prediffuser outlet annulus. This can then have the effect of reducing the amount of diffusion being attempted in the region of the pre-diffuser walls, and increasing it near the duct centre, thereby reducing the axial non-uniformity of the outlet flow. The possibility then exists to use relatively short, wide angle systems, without the occurance of separation, or gross flow non-uniformity. It should however be noted that reducing dump gap to a very small value

41 effectively reduces the pre-diffuser outlet area, \flith the result that this 'diffuser' can operate as a nozzle. I The size, and shape of the head is important, since as well as influencing the upstream flow, it largely governs the amount of turning which must be accomplished, and the way in which it takes place. The significance on overall performance of the division of flow into the two settling length annuli can be clearly seen by referring to Equation , and Figure 1.4.1, in which it can be seen that the ideal pressure recovery maximises at a flow split, the design value, at which the mean velocities in the two settling length annuli are equal. Flow split influences pre-diffuser flow, since pressure gradients generated by the flow curvative around the head modify the pressure field at outlet of the pre-diffuser REVIEW OF PREVIOUS WORK SUMMARY OF WORK BY FISHENDEN(l) A brief summary only of this work appears in this section, since much of it can be directly compared with the present work, and therefore appears in the discussion. Tests were conducted on a fully annular dump diffuser rig, similar to that used in the present investigation, with a design flow split of 2.15, and overall area ratio of 2.0. The geometry of the system is presented in Figure Experimental performances evaluation was conducted over a wide range of flow splits, dump gaps, and pre-diffuser geometries, with fully developed flow presented at entry to the system.

42 - 2'1 - It is now convenient to view the main results obtained for the pre-diffuser, and overall system separately. '1.6.'1.'1 THE PRE-DIFFUSER (a) It was found that the majority of change of prediffuser pressure recovery with dump gap or flow split was due to insufficient and not inefficient diffusion i.e. resulting from a change of the non-uniformity of outlet veloci ty profiles. (b) For any given dump gap and pre-diffuser geometry, an optimum flow split existed. This optimum point was simi- 1ar whether maximum pressure recovery or minimum loss was used as an optimising criteria,' and although influenced by both dump gap and pore-diffuser geometry, always occurred at a flow split below the design value of 2.15, often close to a value of about 1.3. Optimum performance was found to be related to a symmetric pre-diffuser outlet velocity profile with minimum axial distortion (i.e. minimum 0(2). The use of a pre-diffuser, canted outwards at , did however bring the values of design and optimum flow splits closer together. (c) The effect of reducing dump gap was to increase the influence of flow split on the pre-diffuser flow, and also, particularly in the region of the optimum flow split, to improve the unformity of the flow OVERALL (a) For a given pre-diffuser and dump gap, the flow split giving maximum overall performance was generally of a

43 higher value than that giving maximum pre-diffuser performance, although lower than the design value of 2.15 in most cases. (b) Since, at all conditions, there was no gross non-uniformity of the settling length profiles, the reduction of pressure recovery below the ideal value was prima-. rily attributable to loss. (c) No absolute optimum operating condition could be defined in view of conflicting requirements of minimum length, and maximum performance. The overall length could be decreased by: (i) Decreasing pre-diffuser area ratio for a fixed wall angle (ii) Decreas.ing dump gap Uii) Increasing pre-diffuser included angle (2!2l) for a given area ratio. Each of the above however had adverse effects on performance, and pre-diffuser flow non-uniformity. (d). As a general guide only, the division of loss within the system was as follows: Pre-diffuser 25% Dump region 15% Annulus surrounding flame tube (3-4 in Fig ) 60% ISOLATED ANNULAR DIFFUSER PERFORMANCE Considerable work has been carried out by Sovran and Klomp(2) on the testing and performance evaluation of a large number of diffuser geometries. Although, for the system used in the current investigation, the head influences the prediffuser flow, this effect is minimised at large dump gaps../

44 This is clearly seen from Figure 1.6.1, in which it can be seen that good correlation can be obtained betwe~n the results pf an isolated diffuser, and the pre-diffuser at large dump gaps, each having similar entry conditions THE INFLUENCE OF ENTRY CONDITIONS MACH NUMBER AND REYNOLDS NUMBER The influence of Mach number and Reynolds number relating to diffusers have been established from previous work. (i) Little and Wilbur(3) have shown from tests on conical diffusers, that pressure recovery is essentially independent of Mach number, provided that sonic conditions do not prevail in the I (H) critical regiqn of the inlet corner. (4) McDonald and Fox have shown, also in connection with conical diffusers, that performance is essentially insensitive to changes of Reynolds number above a value of about 10 4 (iii). Gurevich(5) has shown that for low entry swirl, annular diffuser loss coefficient is unchanged by Mach number variation between values of 0.25 to 0.7. It is assumed that these results apply equally well to the branched system of the current investigation ENTRY VELOCITY DISTRIBUTION AND TURBULENCE CHARACTERISTICS In view of the difficulty of changing the shape of a velocity profile without altering its turbulence structure,

45 .1.., isolation of the individual effects of each is also difficult. Considerable work has been conducted on the effect on diffuser performance of changing the entry velocity profile, maintaining the turbulence level as near constant as possible, and assuming that all major effects are primarily due to changes of velocity distribution. In general, the effect of distorting an entry flow, either in the direction of flow, or normal to it, is to reduce the pressure recovery attained. In the absence of separation, this is primarily due to insufficient,rather than inefficient diffusion. This is the result of the tend- _ ency to accentuate'flow non-uniformity in a positive pressure gradient, as can readily be seen from the Navier-Stokes equation for steady two dimensional incompressible flow. In the x direction: ~u v,' - ay It is evident, by consideration of the first terms on left and right hand sides of this equation, that in a diffusing flow, where ~ is positive, the reduction of velocity in the x velocity. dx direction is inversely proportional to the local Mixing will modify this result, since it has the effect of tending to make the flow more uniform by redistribution of fluid. Increase of velocity profile non-uniformity implies an increase of kinetic energy, thereby decreasing the pressure recoyery possible in a diffuser. Generally, the result of increasing the turbulent mixing

46 of the entry flow is to improve the pressure recovery, and delay separation, by re-energising the low velocity regions near the walls, and thereby producing a more uniform velocity I profile at outlet. A small increase of loss is usually incurred. Extensive work on the correlation of the effects of presenting variously distorted entry velocity profiles have been carried out by'sovran and Klomp(2), using the blockage factor, B 1, (see Equation ) as a measure of inlet nonuniformity. An empirical relationship between two-dimensional effectiveness, E, II; outlet effective area fraction, E 2, (Equation ) geometric area ratio, AR; and entry blocked area fraction, B 2, was obtained,.as shown in Figure In this way, comparison of the data due to Tyler and Williamson (6), Wolf and Johnston, and Sovran and Klomp(2) is possible, as presented in Figure Investigations into the effect of changing entry mix-. (8) ing have been undertaken by Bradley and Cockrell,and Williams(9). The results indicate that by doubling the intensity of turbulence near the wall of a fully developed entry velocity profile, a 10-12% improvement of pressure recovery could be obtained ENTRY SWIRL The effect of entry swirl on various annular diffuser geometries has been investigated by Gurevich(5). It may be seen in Figure that swirl is generally detrimental to performance, except for diffusers of large wall angle, where an improvement is possible over a limited range.

47 The work of Horlock(10) showed the variation of flow pattern as swirl is introduced, as shown in Figure It can be seen that the separated region moves from outer to inner diffuser wall as considerable swirl is introduced. Hence, at moderate degrees of swirl, the possibility exists to eliminate separation entirely THE AERODYNAMIC STABILITY OF A BRANCHED DIFFUSER A theoretical investigation by Ehrich(11) has been carried out in order to determine a stability criterion for a branched diffuser system, and is briefly presented below. (i) (ij) Let the flow be disturbed from an equilibrium state, (i), to the condition shown in (ii). For static stability, the initial reaction should be to restore equilibrium, and therefore when disturbing the flow, pressure in the inner annulus must rise to a greater extent than that in the outer falls, in order to provide a pressure force which will tend to restore the original division of flow.

48 Hence 6pi > ~po for static stability or ~CPi > 6C po but '" g-c pi = 'oc pi OQi &Q &'C po V' -DC o Qo =.~ &Q '" V\ '0 c pi &Q < -'0 C bq po '0 Qi oqo + o THE CHOICE OF SYSTEM TO BE INVESTIGATED. (1) In the light of recent work by Fishenden,the importance of matching the pre-diffuser, and downstream sections has become clear. It has been shown in this work that optimum performance is related to: (i) A symmetric pre-diffuser outlet velocity profile (ii) A symmetric static pressure distribution over the head, and pre-diffuser outlet annulus (iii) A flow split less than a design value of It can be seen from Figure that the overall ideal pressure recovery maximises at the design flow split, and hence an improvement of performance could be expected if optimum performance also occurred at this point. With this in mind, the design flow split was changed to a value

49 of 1.2. This was based on the fact that this is approximately the ratio of outer to inner mass flows of a symmetric pre-diffuser outlet profile, divided about its centreline. In order to be able to directly compare the results of Fishenden(1), and those of the present investigation, the main parameters, such as head shape, size and position, overall area ratio, and pre-diffuser geometries, remain unchanged, the alteration of the design flow split being accomplished by moving the inner and outer casing walls. (see Figure 2.1.3) 1.8 THE SCOPE AND AIMS OF THE INVESTIGATION The objective of this work is to obtain a better understanding of the fluid dynamic behaviour of a combustion system dump diffuser, to provide details of performance over a range of operating conditions, and to suggest possible improvements. With these points in mind, the following tests have been conducted: (i). Performance evaluation over a range of dump gaps, and flow splits, with each of three axisymmetric pre-diffusers of 12 0 included angle, and area ratios of 1.4, 1.6, and 1.8 respectively. These tests were conducted with fully developed entry conditions, and a design flow split of 1.2 (ii) Performance evaluation over a range of dump gaps and flow splits, with a single 1.6 area ratio, axisymmetric pre-diffuser of 12 0 included angle, and a distorted entry velocity profile, with a design flow split of '

50 In isolation, the results of these tests show: (i) The effect of a variation of pre-diffuser area ratio at a fixed included angle (H) The result of variation of dump gap (Hi) The result of changes of flow split. When compared with the resul ts of Fishenden'( 1), they also show: (i) The effect of changing entry conditions at various common downstream conditions (ii) How a change of design flow split affects performance and fluid dynamic behaviour.

51 SECTION 2 THE EXPERIMENTAL FACILITY 2.1 CONSTRUCTION OF THE TEST RIG The general arrangement of the test facility can be seen in Figures and 2.1.2, detailed dimensions being presented in Figure A fully annular system has been used, in view of the uncertainty of the end wall effects associated with segmented models. Flow was supplied by drawing air through the rig by means of a Keith and Blackman 2513S centrifugal fan, driven by a D.C. electric motor, fitted with resistive speed control. The use of a suction system, with remote atmospheric exhaust, and the presence of a plenum chamber and honeycomb between fan and test regions, minimised the effect of the fan flow characteristics on the airflow through the rig. In order to prevent large scale ambient air movements from having a significant effect on the flow through the rig, honeycomb was incorporated into an 8:1 contraction ratio entry flare. Additionally, trip wires were attached to both inner and outer annulus walls just after the start of the parallel entry length, in order to ensure stable, and circumferentially uniform transition to turbulent flow. Fully developed flow was then ensured at the end of the 24 hydraulic diameter entry length. By using a vertical construction, the use of support struts was minimised, these only being found necessary at the start of the entry length, and in the downstream "regions of the settling length. Clean flow was therefore ensured in the sections in which measurements were taken.

52 Perspex was chosen as the main structural material in view of: (i) The relative ease of manufacture (ii) The high accuracy possible (typically ~ 0.07 mm) (iii) The possibility of using flovl visualisation techniques (iv) The need to 'set up' traverse probes. 2.2 TEST RIG VARIABLES (i) Dump Gap The combustion chamber assembly was mounted on a screw jack, enabling it to be moved vertically by means of a calj,brate.{.. wheel mounted at the base of the rig. In this way, the dump gap, D', could be varied between 0 and mm with an accuracy of _ 0.1 mm. (ii) Flow Split The variation of flow split was facilitated by vertical movement of a profiled throttle ring mounted at the end of the settling length outer annulus on three equispaced lead screws, as shown in Figure In order to extend the range of flow splits available, a fixed throttle of 66% area blockage could be fitted to the end of the inner settling length annulus. (iii) Pre-Diffuser Geometry A range of pre-diffuserswith various geometries was available, these being interchangeable without modifications to the test rig. The dimensions of those used in this series of tests (designated 1, 2 and 3) are given in Figure

53 It can be seen from Figure that these geometries are (2, close to the optimum lines of Sovran and Klomp '. 2.3 INSTRUMENTATION (i) Pre-Diffuser Entry and Exit Facili ties for emp.loymg the manual traverse mechanism shown in Figure were located at three circumferential position, mutually at 120 0, at stations 1 and 2, as defined in Figure Entry traverse positions were located two annulus heights upstream of the start of the pre-diffuser in order to provide invariant entry reference conditions, with a uniform static pressure. The miniature pitot and wedge static probes used to obtain measurements are shown in Figures and (ii) Combustion Chamber Head In view of the movement of the combustion chamber necessary to vary dump gap, it was not feasible to locate a traverse mechanism on the head. Fixed rakes were therefore employed, -located on the inner and outer head regions at 30 0 to the horizontal, the details of which are given in Figure Additionally, calibration checks on the outer head rake were possible, at a single value of dump gap, by employing the single traverse facility located on the outer wall of the rig. (iii) Settling Length Traverse In view of the radial uniformity of the static pressure in the settling length annuli, and the circumferential symmetry of the flow, it was found necessary to conduct total pressure traverse only, at a single circumferential -

54 location. Due to the inaccessible nature of the inner annulus, it was necessary to use the traverse system shown in Figure (iv) Wall Static Tappings Static pressure tappings were provided on all test rig walls as shown in Figure These were located at three circumferential positions, each having a diameter of 0.79 mm. (v) Approximate Flow Split Determination In order to assist in setting the throttles to obtain a given flow split, instruments giving an approximate mean total pressure in _the settling length annuli were fitted. These consisted of lengths of tubing, mounted radiaily across each annulus, blanked off at one end, with several small holes drilled into them,' facing the direction of the oncoming flow. (vi) Recording of Pressure s Pressures from the instruments described above were fed, via plastic pressure tubing, into a Furness Micromanometer, the output signal of which was recorded on a D.I.S.A. type 55 D 30 digital voltmeter.

55 _ 34 - SECTION 3 EXPERIMENTATION 3.1 SCOPE OF TESTS For each of three pre-diffusers, tests were conducted over a range of dump gaps and fldl1 splits, with an overall design flow split of 1.20, and fully developed entry conditions. Additionally, the effects of distorting the entry profile were investigated in tests conducted over a range of, flow splits and dump gaps with a single pre-diffuser, and design flow split of The range of these tests is summarised in Figure 3.1.1, the values of dump gap, and flow split having been chosen on the basis of previous experience so as to include the regions of greatest interest. Each test included measurement of the folloyling items: (i) Total and stdtic pressure profiles in the prediffuser outlet plane (station 2) (H) Total and static pressure data from inner and outer head rakes (stations 3. ~ and 3 ) 0, (Hi) Total pressure profiles in both settling length annuli (stations 4i and 4 0 ) (iv) (v) Static pressures from the wall tappings 'Key' static pressures from wall tappings. (i), (ii), (iii) and (iv) above were conducted at a single circumferential location, (v) being conducted at each of the three circumferential locations available. Each test has been designated a number for convenience of reference, the system used being explained by the use of the following example:

56 Pre-diffuser reference number TEST NO I Non-dimensional dump gap (D/h2 = 0.5) Approximate flow split (S = 1.8), In addition, complete blocks of tests can be referred to: e.g test series refers to the range of tests conducted with pre-diffuser 2, and dump gap ENTRY CONDITIONS (i) Fully Developed Entry Profile Total and static pressure profiles were obtained by means of traverses at station 1, at each of three circumferential locations. These were conducted at extremes of flow split and dump gap outside the normal testing range. Circumferential uniformity, independence of downstream conditions, and the absence of a radial static pressure gradient were ascertained, and the entry velocity profiles presented in Figure were assumed to apply for all test conditions. (ii) Distorted Entry Conditions A perforated ring, as shown in Figure 3.2.2, was used to produce a velocity profile distorted to~jards the outer wall at entry. This system was chosen in an attempt to limit the increase of mixing presented to the pre-diffuser, bearing in mind construction and mounting limitations. Total and static pressure profiles were conducted at station 1 for various positions of the ring, and a location.'

57 providing a velocity profile typical of that at compressor exit was established. The resultant velocity profiles, and also shear stress distributions obtained by using hot wire anemometry, are presented in Figure It can be seen that the closest approximation to typical compressor exit conditions is when.the ring is mm from the entry station, and it was therefore at this condition that testing was conducted. During all tests, the entry velocity was maintained approximately constant at a mean value of about 26 m/s, corresponding to a Reynolds number t1 2h1) 5 of 1.6 x 10 If 3.3 EXPERIMENTAL TECHNIQUE -. Total and static pressure profiles were obtained using the equipment described in Section 2.3. Traverse reference positions were obtained by moving the probes until just touching the rig walls, enabling, with a knowledge of probe size, the probe location to be known to an accuracy of about 0.1 mm. Traverses were conducted from both inner and outer walls of each annulus, with a region of overlap near the duct centre, the step size between reading being chosen as that consistent with accurately defining the profile.

58 Station Annulus No. of points No. of points height in total in static pressure pressure traverse traverse mm mm mm 25 ~ mm All traverse and head rake pressures were recorded with reference to local wall statics (i.e. P2wo' P9Hi/o' P4./ ), these reference pressures, and all wall static w~ 0 pressures also being recorded, referenced to the wall static pressure at station 1. Additionally, the maximum dynamic head at inlet was recorded, before and after each group of readings. 3.4 ACCURACY By selecting suitable damping values on both the micromanometer and digital voltmeter, mean values of fluctuating pressures could be recorded to within ~ 0.2 mm water, for a value of maximum entry dynamic head of about 50.0 lrun water. However, the pressure probes operated under conditions ranging from steady fully developed flow to separated flow, and at largely unknown incidences, the effect of which is shown in Figure A general assessment of experimental accuracy is afforded by the discrepancy between integrated mass flows at each station:

59 Pre-diffuser outlet + 8", m = 2 m 1 _ 0/0 mean for all tests m % mean for all tests m +2.7% 1 In view the fact that calculated parameters are based on large numbers of. experimental data, it is difficult to provide an accurate assessment of likely error. Realistic estimates of the maximum likely errors in the more important. performance parameters are given below: Pre-Diffuser Overall Parameter Typical Error Typical Error vcllue value Pressure Recovery '" C p (~ 5%) (~ 3%) Loss '" t- - Coefficient (.! 40%). (.! 10%) '" wi th (' "'1_2 It should be noted that the high error associated is the result of the low value of pre-diffuser loss, calculated from the difference of large quantities.

60 SECTION 4 PRESENTATION AND DISCUSSION OF RESULTS In this section, the performance characteristics, and fluid dynamic behaviour of the system will be presented and discussed, with particular regard to variation of flow split, dump gap and pre-diffuser geometry. In Section 4.1, the system is viewed as a whole, and in subsequent sub-sections, the flow in various regions of the system is analysed in more detail OVERALL PERFORMANCE (1-4) OPTIMUM CONDITIONS The overall design objectives may be stated as follows: (i) To obtain a maximum pressure recovery (ii) To incur minimum total pressure loss (iii) To minimise the overall length (iv) (v) Good stability, and radial and circumferential flow uniformity To satisfy the above four points at both on and off-design flow splits. The overall performance for various flow splits, dump gaps, and pre-diffusers is presented in Figures 4.1.1;3 'and compared with the data due to Fishenden (1 ). By taking, for given dump gaps and pre-diffusers, values of maximum pressure recovery, and minimum loss, it is possible to obtain curves such as those shown by the unbroken li.nes of Figure A tangent to these curves (the broken line) can 'then be drawn, giving, for fixed overall length, the maximum possible pressure recovery, and minimum pussible loss. It can then be

61 seen that the first three design objectiyes are not consistent. When the overall length is low, pressure recovery is poor, and loss high. However, as non-dimensional length (L /h 1 ) is increased up to a value of about 5.5, an im~rove~ ov _ ment in the maximum attainable performance is possible, although it falls gradually as this length is exceeded. If therefore minimisation of length were not a prime consideration, to obtain the best possible performance for this type of system, it should be operated with a non-dimensional length of 5.5, corresponding to a dump gap (D/h 2 ) -of 1.1 (D/h = 2.0), a pre-diffuser length (L/h 1 1 ) of 3.5, and a flow split close to the design value of 1.2. In this case, an overall pressure recovery of 0.58 may be obtained, with a loss coefficient of Since optimum flow splits occur near the design value, at optimum conditions, the ideal inner and outer pressure recoveries are approximately equal. Loss however modifies this result, and equal"lty of inner and outer pressure recoveries invariably occurs at a flow split above design, as indicated by the example of Figure 'The result is, that at optimum conditions, the pressure recovery of the outer flow field is below that of the inner, as shown in Figure Further curves such as those of Figure may be drawn giving the maximum performance at flow splits other than design. It is then possible to derive charts which enable systems of similar geometry to be designed, and the performance at off design flow splits to be predicted. These

62 are presented in Figures 4.1.7/9, in whi7h the maximum overall pressure recovery, and minimum loss attainable for various overall lengths and flow splits are presented. Lines of constant dump gap have been superimposed such that at any point, the pre-diffuser length may also be determined. It should be noted that, for convenience, dump gap has been non-dimensionalised by h1' and not, as is more usual, by h 2 Since it is usually desirable to avoid gross flow non-uniformity, the regions in which pre-diffuser separation is likely to occur rove been indicated. It should be noted that it was not fou~d possible to construct minimum loss design curves for flow splits in ~xcess of 1.2 because of the relatively small variation of loss coefficient in this region THE INTERRELATION OF PRE-DIFFUSER AND OVERALL PERFORMANCE In order to study the extent to which the pre-diffuser and overall performance characteristics are interrelated, it is helpful to isolate the variables, flow split, dump gap, and pre-diffuser length. Typical variations of pressure recovery and loss coefficient with these parameters, both overall, and within the pre-diffuser, are presented in Figure For a given pre-diffuser and fixed flow split, the beneficial effect of small dump gaps on pre-diffuser flow can be seen, pre-diffuser pressure recovery rising as dump gap is reduced, at the expense.of only a small increase of loss. However, reduction of dump gap increases loss downstream of the pre-diffuser fo such an extent that the variation of overall pressure recovery is the complete reverse of that in the pre-

63 diffuser. It is also clear from this graph, and indeed all those of Figure , that the majority of the pressure rise occurs within the pre-diffuser, little or none occurring downstream due to the high loss in this region. The effect, on performance, of variation of flow split at a fixed pre-diffuser geometry and dump gap is shown in Figure Pressure recovery falls, and loss rises as flow split departs from the design value. This variation is particularly apparent 'in the overall flow, especially the change of pressure recovery, since even under ideal flow conditions, a similar variation of pressure recovery would occur. (see Figure 1.4.1),An example of the effect of changing pre-diffuser length (and hence area ratio since 2~ is fixed) at constant values of dump gap and flow split, is presented in Figure It is clear, that as pre-diffuser length is increased, more diffusion is being attempted, and a higher pressure recovery could be expected within the pre-diffuser, albeit at the expense of an increase of loss. In this way, it is therefore possible to modify entry conditions to the region downstream of the pre-diffuser, with the result that as more diffusion occurs within the pre-diffuseg loss downstream of it, falls. However, for the particular example shown, a gain of overall performance in this way is only possible by increasing nondimensional pre-diffuser length up to a value of 3.4. Beyond this value, pre-diffuser loss increases significantly, outweighing any reduction of loss downstream. "

64 FLOI'I STABILITY The stability parameter ~ (as defined in Section 1.6.4) is plotted in Figure against flow split for various values of dump gap. Although the system is stable within the range of the experimental results, the margin of stability changes with both dump gap and flow split. In the region of the design flow split, the system is theoretically least stable, although greater stability may be achieved at the smaller dump gaps THE INFLUENCE OF DISTORTED ENTRY CONDITIONS It has been shown in Figure 3.2.'3 that, a considerable increase of turbulent mixing was introduced into the flow when the entry velocity profile was distorted. The consequence of this was an increase in overall loss, and a reduction of pressure recovery at all values of flow split and dump gap, as shown in Figure In view of the need to turn a considerable amount of the flow outwards in the dump and head regions at high flow splits, it might be expected that under these circumstances, improvement of performance could be possible for an entry flow already distorted outwards. Since no improvement was in fact measured, it must be assumed that the entry mixing characteristics have a greater influence than the velocity profile shape.

65 i PRE-DIFFUSER PERFORMANCE (1-2) THE INFLUENCE OF DOI'/NSTREAM CONDITIONS ON PRE-DIFFUSER FLOVI In order to provide forces to balance the centrifugal effect as flow turns around the head, pressure gradien~will exist downstream of the pre-diffuser. The influence of these gradients on flow in the pre-diffuser will have a greater influence at small dump gaps, by the very nature of their proximity. Near the design flow split of 1.2, the static pressure distribution around the head is symmetrical about the combustion chamber centreline. The inner and outer boundary layers of the'pre-diffuser flow are therefore influenced by similar pressure gradients, and, when a symmetric velocity profile is presented at entry, the pre-diffuser outlet profile is also symmetric. Under these conditions, the pre-diffuser outlet static pressure distribution is as shown in Figure 4.2.1, with a high pressure prevailing at the duct centreline, although, as dump gap is increased, the pressure variation becomes less apparent. The resultant effect on the pre-diffuser flow can be seen from the outlet velocity profiles of Figure 4.2.1, in which flow uniformity across the duct increases as dump gap reduces. As flow split changes from the design value, the head pressure field will become asymmetric, hence creating an asymmetric pressure gradient at pre-diffuser outlet. Consequently, as a result of the differing axial pressure gradients ( d pi 'Ox' influencing inner and outer pre-diffuser wall

66 boundary layers, the outlet velocity profile is distorted, as shown in Figure PERFORMANCE NEAR DESIGN FLOW SPLIT Since total pressure loss is an inevitable consequence of turbulent mixing, in order to minimise loss, it is normally desirable to mi.nimise mixing within the pre-diffuser flow. However it should be noted that an increase of mixing can eliminate separation, and the high loss associated with it. This implies that the exit velocity profile should be similar to that at entry, although to obtain maximum pressure recovery, not only should loss be minimised, but the outlet velocity profile must be uniform, since in this case, the maximum amount of kinetic energy will have been converted to static pressure. However, since it is not usually possible, in diffuser flows, to obtain an outlet profile more uniform than that at entry, the conditions for minimum loss, and maximum pressure recovery are normally indentical, and consistant with those giving maximum outlet flow uniformity. Optimum pre-diffuser performance could therefore be expected to occur at conditions where the kinetic energy parameter, 0(2' is a minimum, and for symmetric, fully developed entry conditions, when the radial distortion parameter, RD, 2 is equal to zero. As can be seen in Figures , these conditions are satisfied, for all pre-diffusers and dump gaps, at flow splits close to the design value of 1.20, further reduction of 0<2 being possible by decreasing dump gap. Reference to Figures confirm that optimum performance does in fact occur at these conditions. The beneficial effect

67 of decreasing dump gap near the design flow split can be more clearly seen in Figure 4.2.9, in which it is also apparent that the effect of changing dump gap has a greater influence at small values of D/h 2 It is of interest to analyse the relative magnitudes of the two contributions to reduction of pressure recovery below the ideal value, namely loss,.and flow distortion. Figure shows that both factors are of considerable importance PERFORMANCE AT OFF DESIGN FLOW SPLITS -.. At flow split other than the design value, the prediffuser will come under the influence of an asymmetric radial pressure gradient, as discussed in Section Radial, distortion of the pre-diffuser flow results, and loss increases as a result of extra mixing. Figures 4.2.3~~show that the reduction of pressure recovery is greater than that attributable to increased loss alone. This is the result of increased kinetic energy associated with outlet velocity profile distortion, as is shown by the variation of the kinetic energy parameter, 0(2' in Figures At small dump gaps, the radial pressure gradient downstream of the pre-diffuser is not only greater in magnitude, but can also have a greater influence on the pre-diffuser flow by the very nature of its proximity. Hence, although a reduction of dump gap has beneficial effects on the pre-diffuser flow near the design flow split, the performance reduction at off design flow splits is greater. This can be seen in Figure 4.2.9, in which, the reduction of pressure recovery resulting

68 from operating at off design flow splits increases as dump gap is reduced THE INFLUENCE OF DESIGN FLOW SPLIT It has already been stated in Section that, when operating with a settling length geometry having a design flow split of 1.2, optimum pre-diffuser performance occurs at a flow split consistent with a symmetric head static pressure distribution and pre-diffuser outlet velocity profile. The results of Fishenden(l) also indicate a similar relationship, al though these conditions do not prevail at the design floyl split of 2.15, but at values close to the design (or optimum) conditions of the present investigation, as shown in Figures Hence, not only are conditions for good pre-diffuser performance confirmed, the independence of the downstream pressure gradient influencing pre-diffuser flow, and design flow split, is also shown. It appears that, even at optimum flow splits, that reduction of.the design flow split from 2.15 to 1.2 can marginally increase pressure recovery by reducing loss. However, this result must be treated with caution in view of the slight differences in entry conditions (Figure 3.2.1) THE INFLUENCE OF PRE-DIFFUSER GEOMETRY of 12 0 Since all pre-diffuser used in the current \~ork were included angle, those of greater length had a larger area ratio, and therefore a greater pressure rise was theoretically attainable. However, the increased loss, and higher

69 flow non-uniformity that results from more diffusion being attempted partially nullifies any increase of pressure recovery. Furthermore, as diffuser length is increased, separation may occur, creating additional loss and flow non-uniformity which result in further reduction of static pressure recovery. The separation limits for various pre-diffuser lengths, dump gaps, and flow splits are presented in Figure , derived from the assumption that separation was imminent when the non-dimensional velocity (u/u) near the wall fell to a value of THE INFLUENCE OF MODIFIED ENTRY CONDITIONS As shown in Figure , the effect of distorting the entry velocity profile,' at all but very low flow splits, is to decrease pressure recovery, due mainly to an increase of loss. It can be seen from Figure , that the kinetic energy of the pre-diffuser outlet flow (0(2) is higher when distorted conditions are presented at entry. This does not necessarily imply a reduction of pressure recovery, since the kinetic energy of the entry flow is also higher. This is clearly shown from Equation , in which the significance of the ratio can be seen V' C pl-2 = 1 - ( 1 ) AR l _ 2 2

70 As flow split varies for the distorted entry case, the. outlet velocity profiles largely retain their original form as indicated in Figure , whereas for fully developed entry conditions, considerable variation of the outlet velocity profiles result from changes of flow split. With distorted entry conditions, reduction of flow split results in an increase of momentum in the low velocity region near the inner wall, effectively producing a more uniform profile, and reducing 0(2. However, for fully developed entry conditions, at flow splits below about 1.2, ~2 rises with decrease of 0(2 flow split, and the value of falls substantially below 0<.1 that obtained with a distorted entry. Hence, at low flow splits, despite any increase of loss associated with"the increase of mixing of the distorted entry profile, improvement of pressure recovery is possible, as can be seen below flow spl~ts of about 0.8 in Figure DUMP AND HEAD REGION (2-3) In this section, an attempt is made to outline the factors which influence the flow in this region, and to relate changes of flow split, dump gap and pre-diffuser geometry to variation of fluid dynamic behaviour. Since, at station 3, the flow is completely divided into two flow fields, the characteristics of inner and outer flows will be co"nsidered separately. In Section , a number of methods of defining performance parameters were reviewed. Loss or pressure recovery may be non-dimensionalised by mean conditions existing either at pre-diffuser entry or exit (Equations , ).

71 51 Using these definitions, variation of performance parameters may result solely from variation of flow fraction to either inner or outer portions of the flow due to changes of flow split. However, pre-diffuser outlet conditions can have a-significant effect on the flow between stations 2 and 3, and hence conditions prevailing in the individual flow fields at pre-diffuser outlet have been used to non-dimensionalise performance parameters, as defined in Equation e.g. = '" \1\ P3i - P2i '" q2i It should be noted that, in view of the limited numoer of probes within the head rakes, the uncertainty of flow direction producing probe incidence effects, and the difficulty of accurately defining the extent of the vortex, some degree of inaccuracy of the performance and boundary layer parameters is possible, although general trends will still be apparent PERFORMANCE CHARACTERISTICS A number of attempts have been made to correlate the loss in both inner and outer regions of the flow, in terms of both the amount and rate of diffusion or acceleration being attempted. However, no completely satisfactory correlation was found possible in view of the additional factors affecting the flow, as described in subsequent discussion within this section. The variations of both inner and outer flow field pressure recovery and loss coefficients (defined as in Equation ) with dump gap and flow split are shown, for pre-diffuser

72 number 1, in Figure Similar trends were observed for the other pre-diffusers tested. It is clear that pressure recovery of the inner or outer flow field rises as dump gap is increased, or as the percentage of the entry flow associated with that particular flow field falls FACTORS INFLUENCING FLOW CHARACTERISTICS FLOW TURNING Loss "is generally associated with the process of turning flow in view of the mixing required to redistribute the flow as pressure gradients are generated in order to balance centrifugal forces. It could therefore be expected that loss would rise as either the rate or amount of turning increased. Since the flow must return to the axial direction in the settling length after turning around the head, then at some point along each streamline a point of inflection must occur. Although the position of such a point cannot be defined precisely, it can be seen from the static pressure distributions of Figures and (see Appendix 3 for the complete series of results), that the pressure gradient at station 3 is consistent with that of a flow curvature which is convex with respect to the combustion chamber. Hence the flow must have began to return to an axial direction prior to station 3. The rate at which the flow turns is largely governed by dump gap, since this determines the length available for this process to be completed. It is considerably more difficult to discuss the amount of turning, since this depends not only on dump gap, but also

73 on flow split. The influence of dump gap on pre-diffuser outlet conditions has already been discussed in Section 4.2.1, in which it was shown that, by reduction of dump gap, a more uniform pre-diffuser outlet velocity profile could be obtained. This would have the effect of reducing the mass flux near the duct centre, and increasing it in the region of the walls, and therefore less turning is needed around the head, since, in effect, some turning has already been accomplished within the.pre-diffuser. Flow split also varies the amount of turning, in a way that is also influenced by dump gap. It may be seen in Figure that as flow split varies at small dump gaps, the position of the dividing streamline at pre-diffuser outlet remains almost unchanged, and close to the centre of the annulus., I I I I DIVIDING STREAMLINE J J I /~i HIGH FLOW SPLIT Qo~Qi LOW FLOW SPLIT Qo«Q; SMALL DUMP GAP From the above sketch it can be seen that at a high flow

74 split, in for example, the inner flow field, a greater proportion of the pre-diffuser outlet flow is concentrated in the region of the dividing streamline. Hence more turning! of inner flow is required as flow split increases, and vice versa in the outer flow field. At large dump gaps, there is considerable variation of the dividing streamline location at pre-diffuser outlet with flow split, although the velocity profile at this location does not change appreciably (see Figure 4.3.4). DIVIDING STREAMLINE Q. (' I I I \ HIGH FLOW SPLIT Qo»0, LARGE DUMP GAP LOW FLOW SPLIT Q.«Q; Taking the inner flow field as an example, it is clear that as flow split is reduced, the effective centre of mass of the pre-diffuser outlet flow moves away from the inner wall, necessitating a greater amount of turning around the head..'

75 DIFFUSION AND ACCELERATION OF FLOW As shown in Figure the flow between stations 2 and 3 invariably undergo a net acceleration, or, at large dump gaps and a low flow fraction, a slight diffusion. However, at large dump gaps, the flow can undergo considerable diffusion in the rapid expansion of the dump, subsequently accelerating over the head. This can have an adverse effect on performance due to the mixing loss associated with the diffusion process THE INFLUENCE OF PRE-DIFFUSER OUTLET TURBULENCE STRUCTURE In view of the large shear stress associated with high velocity gradients, it could be expected that the turbulent mixing of a velocity profile would rise in the region of high radial velocity gradient as flow non-uniformity increased. The kinetic energy flux parameter, 0(2' can be used as a measure of flow non-uniformity, and as can be seen in Figure 4.3.4, for the pre-diffuser outlet inner flow field, this parameter rises with flow split. Hence, considering the mixing characteristics alone, it could be expected that loss around the inner head region would rise due to extra mixing as flow split increased, and vice versa in the outer flow MIXING BETWEEN INNER AND OUTER FLOI'I FIELDS Although it has been found convenient to divide the flow into two parts split about the dividing streamline, in the region up to where the flow becomes completely divided.'

76 around the head, this division is arbitrary, with no solid boundary between inner and outer flow fields. Mass transfer between the two flow fields, is therefore inevitable in the presence of turbulent mixing, this being particularly apparent at large dump gaps. In this situation, energy would be transferred to the flow field having the lower velocity in the vicinity of the dividing streamline, decreasing loss in this region at the expense of an increased loss in the other flow field. f DIVIDING /.. u { le <~ i.\ STBEAMLI~ III HIGH ENERGY REGION ",LOWER ENERGY REGION, ENERGY TRANSFER VORTEX Fl.OW Since the majority of the vortex lies vii thin the region 2-3, the energy required to sustain it must constitute part of the loss within this portion of the flow. An approximate estimate of the loss associated vii th the vortex, found by assuming that all of the kinetic energy contained at station 3 is detroyed has shown that in the order of half of the total loss between 2 and 3 could be attributable to the vortex. "

77 THE INFLUENCE OF DIFFERENCES BETVJEEN INNER AND OUTER FLOW FIELD DUCT GEOMETRIES The choice of a design flow split of 1.20, and annular geometry, dictate inner and outer casing walls VJhich are not equidistant from the combustion chamber centreline. The result is, for similar conditions at pre-diffuser outlet, less turning of the outer flow than that of the inner is required. There is also a difference in the way in which diffusion or acceleration takes place. At large dump gaps it has alr.eady been stated that the flow diffuses into the dump region prior to an acceleration over the head. In the inner flow, the turning procedure around the head implies that the flow is progressing towards a smaller radius, and any diffusion must take place radially, i.e. in the plane of the main velocity profile non-uniformity. However, in turning, since the outer flow is progressing towards a larger radius, a. considerable amount of diffusion can occur circumferentially, which is normal to the plane of flow non-uniformity. Viets(12) has shown, in a theoretical investigation, that there can be a substantial difference in the mixing characteristics of flows diffused in different planes, more mixing being associated with diffusion normal to the plane of velocity profile non-uniformity. Solely as a result of this phenomena, it could therefore be expected that total pressure loss would be greater for the outer flow field.

78 DISCUSSION OF HEAD STATIC PRESSURE DISTRIBUTIONS The variation of static pressure on the head is pre- I sented in Figures It can be seen that, despite changes of dump gap, flow split and pre-diffuser geometry, the magnitude and location of the peak pressure, at the stagnation point, remains unchanged. The fall of pressure around the head, and subsequent rise downstream of a point corresponding closely with station 3, is the result of two influences. Firstly, as a result of the flow acceleration prior to diffusion into the settling length, the mean static pressure of the flow will fall. This continues up to a region where the mean velocity is a maximum, static pressure subsequently rising as diffusion occurs. Secondly, the flow turning, as discussed in Section also influences the head static pressure distribution. Due to the necessity to turn the flow around the head, and then return it to an axial direction, the pressure gradient across the annulus surrounding the head will change both in magnitude and direction in the manner illustrated below. FLOW CONCAVE TO HEAD / NEGA T I V E d P.., dy H,, /FLOW CONVEX POSITIVE TO HEAD dp ay.

79 Hence, in the absence of any diffusion or acceleration, the turning of the flow would produce similar head static pressure variations to those shown in Figures SETTLING LENGTH (3-4) GENERAL FLOW CHARACTERISTICS Since the inner and outer flows are completely separate at station 3, and hence almost entirely independent of each other, performance and flow characteristics \~ill be mainly dependent upon local entry conditions (i.e. at station 3). In general, the flow diffuses rapidly into the initial regions of the settling length, subsequently diffusing more slowly in the constant area regions as the velocity profile becomes more uniform due to mixing. Additionally, the flow is turning towards the combustion chamber, becoming axial in the downstream region of the settling length. For reasons outlined in Section 4.3, performance parameters in this section have been non-dimensionalised by local entry conditions, as defined by Equation The variation of performance with dump gap and flow split is typical of that sho_m in Figure 4.4.1, in which the results for prediffuser number 2 are presented. It can be seen that pressure recovery in both inner and outer flow fields falls as the flow fraction to that particular annulus also decreases. The ideal pressure recovery however rises with decreasing flow fraction, and therefore the reduction of pressure recovery is solely the result of a rapid increase of loss.

80 FLOW STABILITY Consider a flow particle of mass Ill, moving in an arc of radius R, with tangential velocity u. - -mass/m u \ In the absence of significant radial acceleration, the centrifugal force must be balanced by a radial pressure force. = m.2e. f dr.2e. dr tu 2 = R Consider the particle displaced to a new radius, Rd. By conservation of angular momentum, if,ud is the new velocity of the particle, then: The centrifugal force exerted by the displaced particle is then given by: 2 mud, = If the local velocity at the displaced position is u 1 ' the local pressure gradient is given by:

81 = f,u 1 Rd I For stability of the flow, the displaced particle must tend to return to its initial position, and therefore the net force must act in the direction opposite to that of the displacement. i.e. m dp1 drd f > mr 2 u 2 Rd 3 u 1 2 > R 2 u 2 Rd Rd 3 u 1 Rd ~ R,u with R. There, for flow st~bility, the product ur must increase It can be seen in Figure 4.2.2, that, for the majo~ity of the flow at station 3, ur decreases with increasing R. It should be noted that it has been assumed that all of the flow " at this station is turning with a centre of curvature corresponding to the centre of the semi-circle which forms the head. This assumption is certainly valid close to the head, but may not be entirely correct elsewhere. Nevertheless, it is not considered that this would alter the conclusion that the flow is unstable, giving rise to a considerable degree of turbulent mixing. The results of Stevens and Fry(15) confirm that turbulent shear,stress of a concave surface boundary layer increases considerably at the flow negotiates a constant area annular bend..'

82 RE-DISTRIBUTION OF FLOW DOVJNSTREAM OF STATION 3 In view of the high degree of mixing associated with the unstable velocity profile at station 3, a rapid re-distribution of flow could be expected downstream of this location. Given a turning arc of sufficient length this would result in a flow distorted away from the combustion chamber, in view of the direction of flow curvature. In addition to the effect described above, we must also consider the diffusion of the flow, which will tend to accentuate the initial distortion. The resultant shape of the velocity profile in the settling length will therefore depend upon the relative magnitude of these two effects, viz. diffusion and curvature. Figure shows that when the flow to either inner or outer annulus is high, the pressure recovery between 3 and 4 is considerable, resulting in velocity profiles in the settling length distorted towards the combustion chamber, as in Figure However, as the flow to either settling length annulus falls, the instability of the head flow becomes more apparent, as shown in Figure 4.4.2, in which, for the inner annulus, the fall of {u/u)r with increasing radius is more apparent, for the majority of the flow field, at a high flow split. The result of this is an increase of mixing at low flows, a consequential rise of loss (see Figure 4.4.1), and rapid re-distribution of the curving flow, producing velocity profiles in the settling length distorted towards the casing walls, as shown in Figure The plane in which the diffusion takes place has a,"

83 significant influence on the mixing characteristics of the flow, as discussed in Section A considerable amount of the diffusion in the outer annulus can occur circumferentially, which is normal to the plane of velocity profile nonuniformity, unlike the diffusion in the inner annulus, which must occur entirely in the plane of non-uniformity. In view of the extra mixing which Viets(12) suggest would be associated with the outer flow, it could be expected that loss 'would be greater than that of the inner flow, although an improvement of flow uniformity in the settling length would result. These effects can be clearly seen from Figures 4~4.1 and DIVISION OF LOSS WITHIN THE SYSTEM Figures show the percentage of the total loss which is attributable to the various regions of the diffuser system. It is clear that the majority of the overall loss occurs downstream of the pre-diffuser, particularly in the region between 3 and 4. Furthermore, except at ver.y low flow splits, when the inner flow field mass flow is high, a significantly greater prop0rtion of the loss downstream of station 3 occurs in the outer flow field. The influence of dump gap, flow split and pre-diffuser geometry on the division of loss between the pre-diffuser, and the downstream regions of the system can also be seen in Figures 4.5.1/3. The adverse influence on pre-diffuser flow of increasing dump gap, changing flow split from the design value of 1.2, or increasing area ratio, is apparent, since in

84 all cases, the percentage of the total loss which occurs within the pre-diffuser rises.

85 SECTION 5 CONCLUSIONS AND RECOMMENDATIONS 5.1 CONCLUSIONS Low speed tests have been carried out to investigate the performance of a dump diffuser system of overall geometric area ratio 2.0. The system was tested over a range of flow splits and dump gaps with fully developed entry conditions presented to each of three axisymmetric pre-diffusers of area ratio 1.4, 1.6 and 1.8 respectively, and 12 0 included angle, the overall design flow split of the system being Further tests were conducted using the 1.6 area ratio pre-diffuser, with distor.ted entry conditions, and design flow split of the system In addition to considering the overall performance, total pressure loss, static pressure recovery, and flow uniformity for various regions of the system were determined OVERALL PERFORMANCE The influence of flow split, dump gap, and pre-diffuser area ratio has been established, and the optimum conditions defined. By comparison of the results of this investigation with those of Fishenden(l), the influence of distorted entry conditions, and a change of design flow split for constant overall area ratio and combustion chamber geometry, have been established. The main conclusions are: (i) Optimum performance occurs at flow split close to the design value of 1.20, a dump gap (D/h 2 ) of 1.1, and pre-diffuser length (L/R ) of 3.5. This 1

86 corresponds to a pre-diffuser area ratio of.1.74, and overall length (Lov/R 1 ) of 5.5. At these conditions, a pressure recovery of O.$is obtained, with a loss coefficient of (ii) For any given dump gap, pre-diffuser, and design flow split, the optimum performance was obtained at a flow split of about 1.2. This corresponded to asymmetric pre-diffuser outlet velocity profile, and a symmetric static pressure distribution on the combustion chamber head. (iii) In view of the increased level of turbulence associated with the distortion of the entry profile, it.is not possible to isolate the influence of a modified entry velocity profile alone. However, the overall i~fluence was to increase loss at all operating conditions. (iv) Comparison of results obtained for overall design flow splits of 1.20 and 2.15 clearly indicate the need to match system geometry and design flow split. This should be done in such a way that a symmetric pre-diffuser outlet velocity profile and static pressure distribution is obtained at the design flow split PRE-DIFFUSER PERFORMANCE (i) Optimum pre-diffuser performance corresponds closely with a flow split giving optimum overall performance i.e. S ~1.20..

87 (ii) Reduction of dump gap improved pre-diffuser performance near the design flow split, but the performance penalty associated \'ii th off design flow splits increased DIVISION OF LOSS It has been demonstrated that most of.total pressure loss occurs downstream of the point of minimum pressure on the combustion chamber head. This is due to the high degree of mixing associated with the rapid diffusion of a flow which is inherently unstable because of the shape of the velocity profile and direction of flow curvature. Generally speaking, the overall loss can be divided between the various regions of the system as follows: Pre-diffuser (1-2). Dump region (2-3) Settling length (3-4) % 20% 60% However, the exact numerical values of the loss division depend upon dump gap, flow split, and pre-diffuser geometry. 5.2 RECOMMENDATIONS FOR FUTURE viork (i) Performance has been determined over a wide range of operating conditions for pre-diffusers of 12 0 included angle only. It would be of considerable interest to examine the influence of pre-diffusers of larger included angles, and al~o contoured wall shapes..'

88 (ii) Considerable loss is associated with the diffusion and curvature of the flow around the head. There is scope for a considerable reduction of loss if the turning and/or diffusion could be reduced. With this in mind, it might be of interest to conduct tests with various head shapes. (iii) In view of the current trend towards low pollution gas turbines, there is a tendency to increase the percentage of flow entering directly into the primary zone of the flame tube via the head. Hence in any future work, careful consideration must be,given to the inclusion of head porosity. (iv) It is clear that entry conditions can have a significant effect on diffuser performance. In future work, consideration should be given to the fact that fully developed entry conditions may not give a true representation of conditions occurring in practice. An attempt should therefore be made to test with more representative entry conditions.

89 REFERENCES 1. FISHENDEN, C.R. "The Performance of a Branched Annular Diffuser System" - Pd.D. Thesis, Loughborough University of Technology, SOVRAN, G. and KLOMP, E.D. "Experimentally Determined Optimum Geometries for Rectilinear Diffusers with Rectangular, Conical or Annular Cross Section" - Fluid Mechanics of Internal Flow, Elsevier, LITTLE, B.H. and WILBUR, S.W. MCDONALD, A.T. and FOX, R.W. "Performance and Boundary Layer Data from 12-degree and 23-degree Conical Diffusers of Area Ratio 2.0 at Mach Numbers up to Choking and Reynolds Numbers up to 7.5 x 10 6 " - N.A.C.A. Report 1201, "An Experimental Investigation of Incompressible Flow in Conical Diffusers" - A.S.M.E. Pre-print 65 - FE ~ 25, GUREVICH, D.V. "An Experimental Study of Diffusing Exit Channels in Helicopter Gas Turbines" - Helicopter Power Plant II, Oberongiz, 1959 (M.O.D. Joint Intelligence Bureau, D.S.I. Translation No. 680, 1961). 6. TYLER, R.A. and WILLIAMSON, R.G. "Diffuser Performance with Distorted Inflow" - Proceedings of Symposium on Subsonic Fluid Flow Losses in Complex Passages and Ducts, I.Mech. E. Proceedings Vol. 182, Part 3D,

90 WOLF, S. and JOHNSTON, J.P. I BRADLEY, C.I. and COCKRELL, R.G. viilliams, G. J \ "Effects of Non-Uniform Inlet Velocity Profiles on Flow Regimes and Performance in Two-Dimensional Diffusers" - Report PO-12, Mech. Eng. Dept., Stanford University, "The Response of Diffusers to Flow.Conditions at Their Inlet" Symposium on Internal Flows, University of Salford, Paper 5, "The Influence of Inlet Conditions on the Boundary Layer Growth and Overall Performance of Annular Diffusers" - Ph.D. Thesis, Loughborough University of Technology, HORLOCK, J.H. 11. EHRICH, F.F. 12. VIET,s, H... "Boundary Layer Problems in Axial 'l'urbomachines" - Flow Resear.ch on Blading, Elsevier, "Aerodynamic Stability of Branched Diffusers" - A.S.M.E. Paper No. 70-GT-27, "Directional Effects in 3-D Diffusers" - Aerospace Research Labs./ LE, WPAFB, U.S.A., HOWARD, J.G.H., HENELER, H.J. and THORN TON-TRUMP, A.B. "Performance and Flow Regimes for. Annular Diffusers" - A.S.M.E. Pre-print 67-WA/FE-21, LIVESEY, J.L. "Duct Performance Parameters Considering SpatiallY Non-Uniform Flow" - AIAA Paper No

91 STEVENS, S.J. and FRY, P. I "Measurements of the Boundary Layer Growth in Annular Diffusers" - Journal of Aircraft, Vol. 10, No. 2, pp

92 'Fig.1.3,l TWO TYPES OF COt-lBUSTION CHAt-lBER DIFFUSER SYSTEMS IN CURRENT USE SPUTTERS PRIMARY SECONDARY DILUTlO ZONE ZONE ZONE FAIRED SYSTEt-l - ---' ~f)-./.q;. ~' ---~ ,,... PRE~DIFFUSER ~ \ ~ """ \ tb;;',~ ~ ~,t PRIMARY SECONDARY DILUTION ~ ZONE ZONE ZONE, \f ) "'" -:;:/ ~. -:;/'---' " ", DUMP SYSTEM

93 Fig.l:4.1 VARIATION OF OVERALL IDEAL PRESSURE RECOVERIES AND EFFECTIVE AREA RATIOS WITH FLOW SPLIT 0 9~ ~ :::::'-Ir=:->-<'A""=-""'::: ~I CP,-4(max) : DESIGN FLOW S PLI T: ' 20 <v--'7" AR e' "'" CPl s":1.20 S: 2 15 ~--ll~design FLOW SPLI T: 2 15 ~o OVERALL GEOMETRIC AREA RATI0:2 0 FULLY DEVELOPED INLET CONDITIONS (0<1: ) >- FLOW SPLI T (S)

94 Fi RELATIONSHIP BETWEEN KINETIC ENERGY FLUX COEFFICIENT AND BLOCKAGE FRACTrON 1.6 PRE-DIFFUSER No. AR A o o L-..I- --L- --'- --I_---' , "" t"'_ Fig COMPARISON OF PRE~DIFFUSER AND ISOLATED DIFFUSER PRESSURE RECOVERIE~ / / / / / (2) SOVRAN AND KLOMP (ISOLATED DIFFUSER) P R E~DIFFUSE R 0 0 O

95 Fig.-1."S.2 INFLUENCE OF INLET BLOCt<AGE ON DIFFUSER PER FORMANCE AFTER SOVRAN & }\LOMP. 2 ) I. E ' ~= Best fit for all geometries on ~d7rc;. f- 1 (El Ma El ;:;:;'<: AA ~ /E 0,) "'" i'-, Fig VARIATION OF 01 FFUSER EFFECTIVENESS WITH INLET BLOCKAGE FRACTION. 100r r r----r ~r_--, Annular, M=2 0 BO 60. ',/ ~,. :--. AR(100a/4- WOLF & JOHNSTON(B)/'" Two-di mensional. \' "'- uniform/step shear, jet and wake flow at inlet. TYLER &.WILLlAMSON(7)./ Annular, AR=2.25 \ "- \ \ \ 40 \ \ \ 20~~~~~------~~~~L \~~--~. ~ 00050'01 0'05 0'1 0'5 1'0 INLET BLOCKAGE FRACTION, Bl \ \

96 Fig INFLUENCE OF ENTRY SWIRL ON LOSS COEFFICIENT FOR CONSTANT INNER CORE ANNULAR DIFFUSERS AFTER GUREVIC H(5) Plenum Chamber -----t--- Exhaust to atmosphere \ swirl blades 1OO'~r---~---r--~ ~---.. AREA RATIO = o o MEAN INLET SWIRL ANGLE-Degrees. Fig ; TOTAL PRESSURE CONTOURS IN ANNULAR DIFFUSERS. DATA OF HOADLEY REPORTED BY HORLOCK (13) N I~ ARI~ :? ) >; ;; ; 7 ; ;) )/Kua 5rAT/oN-,\RY;;; )))) NO SWIRL DIFFUSER GEOMETRY: N/LlR,=19.1,AR:4 57 /'V4J",CJ indicates SEPARATION. ;) ; A ; ; 7 ) 7 7 ; > ); 7 ; > ; )';))));; --~~----- FREE VORTEX SWI RL (Swirl angle at A =17 5").'

97 Fig LAYOUT OF TEST FACILITY. '\ ( 11 EXHAUST TOA1MOSPHERE. VIA DUCTING PLAN VIEW FOR DETAILS SEE Fig CENTRAL PILLAR & LEAD SCREW SUPPORTING COMBUSTION CHAMBER.'

98 Fig EXPERIMENTAL FACILITY. KEY (1) Intake Flare (2) Inta ke Bull et (3) Inlet Length (4) Inlet Stati on (1) (5) Pre-ditfuse r (6) Combustion Chamber Head (7) Head Rake (8) External Static Pressure Tappin gs (9) Internal Static Pressure Tappi ng s (10) Plenum Chamber (11) Traverse Gear (12) D I SA Dig ital Voltmeter (13) Furness Control s Micromanometer APPROXI MATE SCALE o 0 5 Metres